search

PolusAI / nyxus

A fully scalable robust image feature extraction library.
MIT License
19 stars 8 forks source link

nyxus.Featurize() appears to switch intensity_image and mask_image names #210

Closed TimMonko closed 4 months ago

TimMonko commented 7 months ago

First off, great library! I am glad I found this, as many of the measurements are useful and unavailable elsewhere (at least easily).

Second, when I pass intensity_names and label_names to Nyxus.featurize(), the pandas column names seem intuitively switched. 

stats = nyx.featurize(
    intensity_images=fo, 
    label_images=label_image,
    intensity_names=['FerroOrange'],
    label_names=['Labels']
)

The 'intensity_image' column is populated with 'labels' and the 'mask_image' column with 'FerroOrange'. I would expect the opposite, even without the specific intensity_name var in featurize, I think of masks as the label as well.

Perhaps this is intentional, and just confusion of my intuition! But even passing nothing to those arguments also leads to the confusing 'intensity_image'='Segmentation0' and 'mask_image'='Intensity0'.

nyxus.__version__='0.7.5'

sameeul commented 7 months ago

Thanks for reporting this. I looked at the code and found the issue. We will push a fix soon.

sameeul commented 4 months ago

Fixed in v0.8.1